How to Connect Spring AI MCP Apps to Claude Desktop with mcp-remote Proxy

Mar 25, 2025

Problem

When I tried to connect my Spring AI MCP server to Claude Desktop, I couldn’t make it work. I had my MCP server running on localhost:3001, but Claude Desktop just wouldn’t see it.

I checked my Spring AI configuration:

spring:
  ai:
    mcp:
      server:
        type: SYNC
        name: my-mcp-server
        version: 1.0.0

The server was running fine. But Claude Desktop couldn’t connect to it.

Environment

Spring AI 1.0.0
Claude Desktop (latest)
Java 21
macOS

What happened?

I was trying to connect Claude Desktop to my Spring AI MCP server. My server uses the Streamable HTTP transport, which is the default for Spring AI MCP servers.

But when I configured Claude Desktop’s claude_desktop_config.json:

{
  "mcpServers": {
    "my-app": {
      "url": "http://localhost:3001/mcp"
    }
  }
}

Claude Desktop just ignored it. No connection, no tools visible.

The problem is: Claude Desktop doesn’t support the Streamable HTTP or SSE transports that Spring AI MCP servers use. Claude Desktop expects MCP servers to use a stdio-based protocol—where it launches the server as a subprocess and communicates via stdin/stdout.

How to solve it?

I found the solution: use mcp-remote, an npm package that acts as a proxy between Claude Desktop’s stdio protocol and HTTP-based MCP servers.

Step 1: Start your Spring AI MCP server

First, make sure your Spring AI MCP server is running:

./gradlew bootRun

# Or with Maven
./mvnw spring-boot:run

Your server should be running on port 3001 (or whichever port you configured).

Step 2: Configure Claude Desktop with mcp-remote

Edit your Claude Desktop configuration file:

macOS/Linux: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "my-spring-ai-app": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "http://localhost:3001/mcp"
      ]
    }
  }
}

Note the -y flag—it’s important! Without it, npx will prompt for confirmation and Claude Desktop will hang.

Step 3: Restart Claude Desktop

Close and reopen Claude Desktop. It will now launch mcp-remote as a subprocess, which connects to your HTTP-based MCP server.

Now when I chat with Claude, I can see my MCP tools are available.

Multiple MCP Servers

If you have multiple Spring AI MCP apps, you can configure them all:

{
  "mcpServers": {
    "app-one": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "http://localhost:3001/mcp"]
    },
    "app-two": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "http://localhost:3002/mcp"]
    }
  }
}

How mcp-remote Works

Here’s what happens when Claude Desktop connects to your Spring AI MCP app:

┌─────────────────┐         ┌──────────────┐         ┌──────────────────────┐
│ Claude Desktop  │  stdio  │  mcp-remote  │   HTTP  │ Spring AI MCP Server │
│                 │ ◄─────► │    proxy     │ ◄─────► │  (localhost:3001)    │
└─────────────────┘         └──────────────┘         └──────────────────────┘
        │                           │                           │
        │  Launch subprocess        │                           │
        │ ─────────────────────────►│                           │
        │                           │  HTTP POST /mcp           │
        │                           │ ─────────────────────────►│
        │                           │                           │
        │  Tool request (stdio)     │                           │
        │ ─────────────────────────►│  Forward via HTTP         │
        │                           │ ─────────────────────────►│
        │                           │                           │
        │                           │  Tool response (HTTP)     │
        │  Tool response (stdio)    │ ◄─────────────────────────│
        │ ◄─────────────────────────│                           │

The mcp-remote package bridges the gap between Claude Desktop’s stdio-based protocol and your HTTP-based MCP server.

Alternative: MCP Jam

If you want a simpler setup without the proxy, try MCP Jam. It’s an MCP client that supports direct HTTP connections:

{
  "mcpServers": {
    "spring-ai-app": {
      "transport": "streamable-http",
      "url": "http://localhost:3001/mcp"
    }
  }
}

MCP Jam supports the Streamable HTTP transport natively, so no proxy is needed.

Client Comparison

Feature	Claude Desktop + mcp-remote	MCP Jam	Goose
Direct HTTP	No (needs proxy)	Yes	Yes
Streamable HTTP	Yes (via proxy)	Yes	Yes
updateModelContext	Yes	Yes	No
Setup complexity	Higher	Simple	Simple

Common Mistakes

I made these mistakes when setting this up:

Forgetting the -y flag - npx prompts for confirmation and Claude Desktop hangs
Wrong URL format - Must include the /mcp path, not just http://localhost:3001
Not starting the Spring AI server first - mcp-remote fails to connect
Wrong config file location - Make sure you’re editing the correct path for your OS

Summary

In this post, I showed how to connect Spring AI MCP Apps to Claude Desktop using the mcp-remote proxy. The key point is Claude Desktop uses a stdio-based protocol and cannot directly connect to HTTP-based MCP servers—you need the mcp-remote npm package to bridge the gap. The configuration is straightforward: npx -y mcp-remote http://localhost:3001/mcp.

Final Words + More Resources

My intention with this article was to help others share my knowledge and experience. If you want to contact me, you can contact by email: Email me

Here are also the most important links from this article along with some further resources that will help you in this scope:

👨‍💻 Spring AI MCP Documentation
👨‍💻 mcp-remote npm package
👨‍💻 MCP Jam

Oh, and if you found these resources useful, don’t forget to support me by starring the repo on GitHub!